UBC Faculty Research and Publications

Characterising the SCUBA-2 superconducting bolometer arrays. Amiri, Mandana; Burger, Bryce; Halpern, Mark; Hasselfield, Matthew 2010-12-31

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


52383-Halpern_SPIE_7741_774106.pdf [ 4.12MB ]
JSON: 52383-1.0107574.json
JSON-LD: 52383-1.0107574-ld.json
RDF/XML (Pretty): 52383-1.0107574-rdf.xml
RDF/JSON: 52383-1.0107574-rdf.json
Turtle: 52383-1.0107574-turtle.txt
N-Triples: 52383-1.0107574-rdf-ntriples.txt
Original Record: 52383-1.0107574-source.json
Full Text

Full Text

Characterising the SCUBA-2 superconducting bolometer arrays Dan Bintley1a, Michael J. MacIntoshb, Wayne S. Hollandb,d, Per Friberga, Craig Walthera, David Atkinsonb, Dennis Kellyb, Xiaofeng Gaob, Peter A.R. Adec, William Graingerc, Julian Housec, Lorenzo Moncelsic, Matthew I. Hollisterd,i, Adam Woodcraftd, Camelia Dunaree, William Parkese, Anthony J. Waltone, Kent D. Irwinf, Gene C. Hiltonf, Michael Niemackf, Carl D. Reintsemaf, Mandana Amirig, Bryce Burgerg, Mark Halperng, Matthew Hasselfieldg, Jeff Hillh, J. B. Kyciah, C.G.A. Mugfordh, Lauren Persaudh a  Joint Astronomy Centre, 660 N. A’ohoku Place, Hilo, HI 96720 UK Astronomy Technology Centre, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK c School of Physics and Astronomy, 5 The Parade, Cardiff University, Cardiff CF24 3YB, UK d Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ, UK e Scottish Microelectronics Centre, University of Edinburgh, West Mains Road, Edinburgh EH9 3JF, UK f National Institute of Standards and Technology, 325 Broadway, Boulder, CO 80305 g Department of Physics and Astronomy, University of British Columbia, British Columbia V6T 1Z1, Canada h Department of Physics, University of Waterloo, Waterloo, Ontario, N2L 3G1, Canada i California Institute of Technology, Pasadena, CA 91125 b  ABSTRACT SCUBA-2 is a state of the art 10,000 pixel submillimeter camera installed and being commissioned at the James Clerk Maxwell Telescope (JCMT) providing wide-field simultaneous imaging at wavelengths of 450 and 850 microns. At each wavelength there are four 32 by 40 sub-arrays of superconducting Transition Edge Sensor (TES) bolometers, each packaged with inline SQUID multiplexed readout and amplifier. In this paper we present the results of characterising individual 1280 bolometer science grade sub-arrays, both in a dedicated 50mk dilution refrigerator test facility and in the instrument installed at the JCMT. Keywords: TES, transition edge sensor, low temperature detector, SQUID, SCUBA-2, submillimeter astronomy  1. INTRODUCTION SCUBA-2 is a new wide-field common user camera, being commissioned at the James Clerk Maxwell Telescope (JCMT) on the summit of Mauna Kea, Hawai’i. Operating simultaneously at two wavelengths (450μm and 850μm), with a 50 sq-arcmin field of view, SCUBA-2 is set to revolutionise submillimeter astronomy. SCUBA-2 will be an unprecedented imaging and survey instrument, capable of mapping large areas of sky to great depth (100 times faster than the original SCUBA instrument it replaces), addressing key questions about the origins of galaxies, stars and planets. The new instrument and re-imaging mirrors were mounted and aligned on the JCMT to schedule, a major engineering success. At 3.5 tonnes and with dimensions of 2.3m x 2m x 2m, SCUBA-2 is by far the largest instrument in the history of JCMT. The engineering challenges and achievements of SCUBA-2 in this regard are described in these proceedings by Craig et al. 1 SCUBA-2 came to Hawai’i with two commissioning grade detector sub-arrays with known flaws and limitations, but of sufficient quality to begin the instrument commissioning. Fully populated the two focal planes will each consist of four 1280 Transition Edge Sensor (TES) bolometer sub-arrays. Early commissioning highlighted two problems that would undermine the performance of future science grade arrays in the instrument: Sensitivity to the motion of the antenna (and 1  d.bintley@jach.hawaii.edu; phone (808) 961-3756; fax (808) 961-6516 Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy V, edited by Wayne S. Holland, Jonas Zmuidzinas, Proc. of SPIE Vol. 7741, 774106 © 2010 SPIE · CCC code: 0277-786X/10/$18 · doi: 10.1117/12.856839 Proc. of SPIE Vol. 7741 774106-1 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  hence the instrument) rotating in azimuth and the 450µm detectors saturated (went normal) on the sky. Both problems have been successfully addressed. Firstly the magnetic shielding on the focal plane units was improved by using a custom single piece welded niobium shield, rather than a number of overlapping niobium panels. Secondly the optical loading on the 450μm focal plane was reduced by narrowing the 450μm pass band filter profile and by eliminating the potential for reflections from the inside of the focal plane covers (done for both wavelengths). In early 2009, the first pair (one of each wavelength) of science quality sub-arrays were completed and tested in a dedicated dilution fridge facility at Cardiff University, UK, before being installed in SCUBA-2 at JCMT. On sky commissioning and a short science observing campaign have followed. The first science results are very promising and will soon be published. The remaining six sub-arrays have now been completed and will be installed in the instrument later this summer (July 2010). The original pair of science grade sub-arrays have also been upgraded with new cold electronics modules (CEM), to correct an adverse thermal loading from 1K, discussed below. In this paper we focus on the dark testing and characterisation of the first pair of science grade sub-array modules both in the Cardiff test facility and installed in SCUBA-2 on the JCMT. Elsewhere in these proceedings Holland et al. 2 provide a full update on the SCUBA-2 project, including the first astronomical results, while Dempsey et al. 3 discuss the on-sky calibration of the instrument.  2. SCUBA-2 DETECTORS 2.1 Sub-array modules Individual sub-arrays are packaged as standalone low temperature modules, connected to their own dedicated MCE (Multi Channel Electronics; room temperature control and readout electronics). This modular approach allows for flexibility. We have been able to test and commission the instrument with various configurations of prototype sub-arrays installed, while detector and MUX development, testing and fabrication have continued in parallel; on occasion with simultaneous cold testing in Edinburgh, Waterloo and Cardiff.  Figure 1. SCUBA-2 science grade sub-array module in a handling jig, ready for unfolding.  The detector array and SQUID multiplexer are fabricated separately on two (actually three, as the detector wafer is two wafers bonded together to provide mechanical support during processing) silicon wafers before joining or hybridising with indium bump bonds.4 The bump bonds provide both thermal and electrical contact between the two layers. The upper detector wafer contains 1280 DC coupled molybdenum-copper TES detectors, in a 40 row, 32 column format. A silicon nitride membrane provides a weak thermal link to the thermal bath.  Proc. of SPIE Vol. 7741 774106-2 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  Each detector pixel includes a resistive heater used to compensate for changes in optical power as the sky background changes, enabling detectors to be operated at a chosen bias point for a wide range of sky powers. The heaters play a fundamental role in the operation of the instrument while observing: •  Detectors are individually calibrated by measuring their responsivity using a ‘fast heater ramp’, (a triangle wave on the heater of a few pW, peak-peak) at the start of and if we choose, periodically during every observation.  •  The optical power from the sky is directly measured by using heater tracking. A servo loop on the heater is run to keep detector output constant while opening and closing the cold shutter from dark to the sky. Periodic heater tracking transfers the slow changes in sky power to the heater setting, keeping each detector close to the original setup operating point in the superconducting transition.  The incorporation of the heater into the design of the detector device and use of the heater while observing is one of the innovative features of SCUBA-2. The 450μm and 850μm detectors have a similar physical design for ease of manufacture, (the 850μm focal plane fully samples an area of sky, whereas the 450μm focal plane is under-sampled by a factor of 2). The transition temperature of the 450μm TES is made higher, to allow for greater power handling and a relaxed NEP requirement. The functional requirement at both wavelengths is that the instrument is sky background noise limited on the telescope. The lower wafer is a NIST designed and fabricated time division SQUID multiplexer (MUX) with 32 channels, each with 41 rows. The 41st row is a ‘dark row’ without any TES element. The dark row SQUID output is available to the data reduction pipeline, to remove excess noise particularly at low frequencies and magnetic pickup. It is the SQUID MUX which makes large-scale TES arrays such as SCUBA-2 practical, by vastly reducing the wire count between the detectors and the room temperature electronics. The operation and design of the time division multiplexer for SCUBA-2 is described by de Korte et al.5  Figure 2. The circuit diagram and schematic drawing of a single TES/MUX element. The electrical and thermal contacts between the detector wafer and the SQUID MUX are made by cold welded indium bump bonds.  On the MUX wafer a first stage SQUID (SQ1) sits below each TES detector element. The coupling between the TES and the SQ1 is by magnetic flux. Current flowing through the TES generates a magnetic field at the SQ1 via an input transformer on the MUX wafer (see Figure 2). Each channel or column of 41 SQ1s is then coupled by a summing coil to the second stage SQUID (SQ2). The signals in each channel are amplified by a 100 SQUID Series Array (SSA) located on the 1K PCB of the cold electronics module (CEM). The SCUBA-2 TES detectors are operated in an approximate voltage biased mode, using a small 5m Ω shunt resistor located on the MUX wafer. The operation of Transition Edge Sensors in this fashion is described by Irwin6 and Irwin and Hilton7, where they summarise the TES small signal theory, applicable to the SCUBA-2 detectors. The advantage of voltage biasing the TES, is that negative electro-thermal feedback (ETF) stabilises the detector against thermal runaway. An increase in background power warms the device and causes an increase in resistance. This in turn causes the detector current to decrease, cooling the TES. Strong ETF essentially keeps the temperature of the TES constant, while providing a simple and direct relation between any applied power (optical or heater) and the current  Proc. of SPIE Vol. 7741 774106-3 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  flowing through the device. Negative feedback also makes the detector self-biasing in terms of temperature in the transition. This is important for large arrays of TES such as with SCUBA-2. If a variation of transition temperatures and transition widths exists across the array and they do not all overlap, the Joule power dissipation in each TES will cause it to self heat to within its respective superconducting transition. Negative feedback can also increase the bandwidth of electrical systems including TES. Changes in the incident power are compensated for by changes in the bias current power on timescales shorter than the thermal time constant of the detector. The current flowing through each TES is measured by its SQ1. The output of a SQUID is periodic with magnetic flux from the input coil; the periodicity being given by a flux quantum. There is no unique output for a given detector current. Therefore we use the first stage SQUID as a null detector. Feedback current is applied by the room temperature electronics (the MCE) to the SQ1 feedback coil to null the field from the TES detector current in the input coil. By applying a flux locked loop (FLL), the applied feedback current is proportional to the current flowing through the TES. We refer to measurements using the flux locked loop as ‘closed loop’ measurements. In open loop measurements we take the output directly from the SQ1, with no feedback applied. The dynamic range of the detector feedback circuit is limited by the available first stage SQUID feedback current and the mutual inductance of the SQ1 input coil. These parameters have been carefully chosen to meet the stringent noise requirements of the instrument. However using the MCE we have been able to increase the dynamic range by flux jumping. The SQUID response is periodic and the period is measured as part of the sub-array set-up. It is possible for the MCE to jump or slip multiple flux quanta, while keeping count of the number jumped and remain locked in the FLL. The dynamic range is increased seamlessly, towards the physical limits of the TES power handling. With too much applied power (optical, thermal, bias) the TES becomes normal and ceases to work as a bolometer, too little and the TES becomes superconducting, to the same effect. 2.2 Upgrading to Science Grade Arrays The manufacture of the SCUBA-2 sub-array is a long process with many steps and testing at each stage. Prior to hybridisation, MUX wafers, which are by far the most complex of the component parts, are individually characterised in a dedicated 4K facility at the University of Waterloo. Testing at this stage allows minor fabrication faults to be identified and corrected. The expected detector yield, based on the uniformity of the first stage SQUID properties is calculated for each MUX wafer.  Figure 3. The results of SQ1 Icmax measurements, done in a dedicated 4K facility, at the University of Waterloo: (Left) 850µm science grade MUX: (Right) 450µm MUX.  The time based multiplexing scheme used requires a common SQ1 bias applied per row, together with a single value of SQ2 feedback per column. When setting up of the sub-arrays, these two parameters are optimised to maximise the number of working detectors. By using the same optimisation algorithm on the Waterloo 4K screening data, we can predict the expected detector yield for a particular MUX. For the latest generation of multiplexer yields exceed 95%. An example of the MUX screening results are given in Figure 3: We show the SQ1 Icmax values, which relate to the SQUID bias currents that give maximum modulation of the V- φ characteristic, for the MUX wafers used for the first science  Proc. of SPIE Vol. 7741 774106-4 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  grade arrays. The gradientss in optimal SQ Q1 bias are an artefact a of the fabrication f process. They makke setting up thhe MUX with a single SQ1 bias per row r and a singgle SQ2 feedbaack current per column more complex. c The detector wafers are carrefully matchedd to the MUX wafers to ensuure reliable bum mp bonding. The T bilayer TES S, heater and associateed ‘wiring’ are fabricated on the underside of o the preparedd detector wafeer; the transitioon temperature is tuned by adjusting the thickness of o the metal layyers. Witness samples s are prooduced during the process, which w are then cooled c to mk temperatuures to verify the t TES transittion temperaturre of each wafeer. The bump boonding process is not withoout risk. We diiscovered durinng the processsing of prototyype arrays, thaat a vital cleaning stepp will aggressiively eat awayy the metal froom the TES devices d if any pinholes in thhe protective layer l are present, channging the propeerties of the TE ES. The 850μm m commissionning array was processed withhout the cleaning step. Unfortunatelyy we found that this left the central regionn of detectors unstable u with different d charaacteristics to thhe edges. This was attrributed to the formation f of ann oxide layer between the twoo halves of som me of the indiuum bump bondds. At the edges of the wafer the deteectors worked as a expected. Here H the bondinng force might have broken through t the oxiide layer producing a contiguous sup perconducting bond. The soluution for the science grade arrays a was to apply a the cleaning step, but only afterr careful and laaborious inspecction of the dettector wafer foor pin holes. Following hybridisation, the t upper deteector wafer iss deep-etch micro-machined m d to reveal the individual detectors d standing on a silicon nitrid de membrane. The T finished detector/MUX d is glued to a beryllium b coppper ‘hairbrush’ thermal and mechaniical support mounting m blockk and then wirre bonded intoo the cold electronics moduule. Room tem mperature electrical cheecks and visuall inspections arre done after eaach step in the process. The design of o the cold eleectronics moduule has provedd very challengging. Over 3000 electrical connections are required between the MUX wafer at a 60mk and a PCB at 1K. The T original solution was too use custom made m superconnducting, niobium traccks on kapton flexes, to linkk the 1K PCB and a ceramicc board (batwiing) with silveer plated coppeer tracks thermally atttached to the hairbrush h mouunt. Aluminium m wedge wire bond links arre made from the silver padds on the ceramic boarrd to the MUX..  Figure 4. Sciennce grade array with w the new styyle cold electronics module, show wing the soldereed connections too the batwing. Inn the expanded view w, the wire bonding, second stagge SQ2s and darkk row SQ1s are visible. v  Testing of thhe prototype and a commissiooning arrays reevealed a weakkness in the wire w bonding process p and a possible heating effecct in some of th he kapton flex to batwing joints. In some cases c the heatinng was sufficient to prevent the subarray from operating. o Wee switched offf those channeels on prototyype sub-arrayss that caused heating, to alllow the remainder off the array to fu unction. The cold electronics modu ule was redesiigned for the science s grade sub-arrays: s Thhe kapton flexees were replacced with t pads on the ceramic batw wing and after many m wire bonnd thermal cyccling and niobium-titannium wire direectly soldered to pull tests, thee wire bond problem has now w been understtood and largelly eliminated. However H testinng of the sciennce grade arrays revealled a new prob blem. The new NbTi wiring introduced i an excessive heatt flow from 1K K to the sub-arrrays and the fridge miixing chamber.. The NbTi wirres are copper clad then insullated. The coppper adds strenggth and allows the wire  Proc. of SPIE Vol. 7741 774106-5 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  to be easily soldered. s To create thermal issolation, the coopper was remooved from eachh wire for a lenngth in the centtre of the cable. The wires were then reinsulated wiith GE varnish,, to prevent eleectrical shorts. A test modulle for new dessign CEM succcessfully cooleed down to 60m mk (at the colld end) in the Cardiff C system m. But as explained beelow; with the 850μm sciencce grade array installed in thhe Cardiff systeem, the fridge base temperatture was held at over 100mk compaared to the usuual 50mk. Thee problem wass traced to the variability of the applicatioon of GE varnish. Thhe GE varnish thickness wass reduced for the 450μm suub-array, whichh reduced the heat leak butt did not eliminate thee problem. The 850μm was thhen repackagedd to match the 450μm 4 before shipping to Haawai’i. A further reddesign of the CEM C has been undertaken. u Thhe cladding onn the niobium-ttitanium wire is i changed to phosphor p bronze, whicch provides thee required therm mal performance without the need for a breeak. The new wire w woven intto ribbon cable has beeen extensively tested at low temperature t at the UKATC and a by the low w temperature group g in Florennce. This new CEM wiill be used for the t new batch of science gradde arrays (incluuding upgradinng the original pair).  3. SUB B-ARRAY TE ESTING The initial tw wo science grrade sub-arrays were tested individually, in a dedicatedd dilution refriigerator facilitty at the University off Cardiff, befo ore careful trannsport and inteegration into SCUBA-2 at thhe JCMT. The final batch of detector arrays will bee integrated dirrectly into the Instrument I witthout further coold testing. The Cardiff test t bed has beeen used to testt a number of prototype p SCU UBA-2 sub-arraays. The magneetic shielding is similar to that on thee instrument, ass is the arrangeement of light tight cold shields surroundinng the detector array. The testt bed has no window too the Lab (alth hough the desiggn is such thatt one could be incorporated);; instead a cryoogenic black body b fills the field of view of the detector d array. The black boody together with w three poinnt source fast switching SPIRE type calibrated illluminators are mounted insidde the 1K shieeld to facilitatee optical testinng. The installled filters havee similar characteristiccs to those in the instrumennt. The dilutionn fridge is connventional in that t there is a liquid He batth and a pumped 1K pot p to condensse the returningg mixture. Hoowever the cryostat is modifiied for optical testing. It has no IVC, so heat swittches instead of o exchange gas g are used for f the initial cool down. Itt is instructivee to compare detector performance in the Cardifff system to thaat in SCUBA--2, to check foor evidence of the pulse tubees, shutter mottor, light leaks etc in thhe instrument affecting a the peerformance of the detectors.  Figure 5. (Leftt) SCUBA-2 sub b-arrays installed in the Cardiff dilution fridge. The T hairbrush suupport is bolted directly to the mixing m chamber, whille the cold electrronics are cooledd by the still. (Riight) Sub-arrays installed in the SCUBA-2 focal plane unit (FPU U) at JCMT.  Proc. of SPIE Vol. 7741 774106-6 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  Previous tests on prototype arrays in the Cardiff test bed have been reported by Woodcraft et al.8 The most significant recent progress in testing sub-arrays has been due to the advances in the use of the MCE. Rather than just characterising individual detectors, we are able to characterise the whole array. This is important because there are many compromises that have to be made to setup an array into the best state for observing. Not every detector can be biased into the optimum part of the transition. Compromises are also made with SQUID setup, which directly impacts the performance of individual bolometers. In previous papers9 we have described the software used for the automatic SQUID setup process. We have successfully automated the process to optimally adjust the 3000 or so parameters that control the MUX operation for the whole subarray in less than 3 minutes. A faster setup, which only adjusts the SSA and SQ2 feedback currents and optimises the SQ1 bias, is used during observing on the telescope. The setup software continues to evolve as we acquire more experience. The quality of the SQUID setup directly affects the performance of the detectors. We observed lower noise from the system with the SSA and SQ2 biased to give maximum V- φ modulation. For the SQ1, we aim to use the minimum bias after optimisation, which has each SQ1 in a row fully on (which tends to coincide with maximum modulation). The bandwidth of the MUX with the SQUIDs biased in this way has been measured using the MCE in the 50MHz fast mode and it is found to be sufficient. The setup procedure itself can be run from the command line or within a script. Scripts can manipulate TES bias or TES heater, taking measurements after each change to simulate applying signals to either line. The instrument fridge temperature is also controllable from the command line or within a script, as is the cold shutter position and external blackbody source. The majority of the measurements to characterise a TES array have been automated in this way. 3.1 Array Characterisation, Basic Tools Using the MCE10, the ability to automate measurement of the TES array is limited mainly by the imagination of the user. At the heart of the MCE are a number of FPGAs. If a particular measurement is not possible with the standard MCE firmware, the FPGA can be reprogrammed to make it so. One example of the MCE flexibility is a fast mode that allows measurements to be made at the 50MHz MCE clock speed. Normally this would not be possible due to limits in the maximum data throughput out of the MCE. In fast mode the MCE stores the measurements, which are then read out of the buffer using the standard MCE command. With fast mode it is possible to investigate the MUX switching and settling times to optimise the multiplexing frame rate and other FLL timing parameters. Another example is an MCE generated heater or TES bias square wave function, used to measure detector time constants. However the most useful and commonly used measurement scripts are quite simple: •  ramp_bias: This is a script which generates a small triangle wave on the TES bias or heater.  •  ramp_bias_vi: This script ramps the TES bias over a (user selected) much larger range, to generate IV characteristics. Usually the ramp is from the normal state into the transition.  •  mceframetest: Collects a time stream of SQ1 feedback current (or direct output if open loop) from all detectors in the array at a selected frame rate (up to a few kHz, with the firmware version we use). By combining a measurement of responsivity (ramp_bias using the heater) with a current noise measurement, the NEP of each detector is measured.  A slightly more sophisticated script is used to measure each detector’s dynamic thermal conductance (G=dP/dT) and superconducting transition temperature, using an open loop, ramping heater method. The script simultaneously measures the heater power required to put each individual detector at the top or normal end of the superconducting transition with minimal TES bias applied. This is repeated for a range of different temperatures. We operate the MCE in two modes; engineering mode for setup and array characterisation and science mode for observing. However the distinction between the modes is reducing, as many of the tasks that previously required engineering mode, can now be done in science mode. The fundamental difference between the modes is that in science mode the MCEs are fully integrated into the telescope control and observing system. The output from the multiple MCEs  Proc. of SPIE Vol. 7741 774106-7 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  are controlled by an external sync box which supplies a clock signal and requests data frames at 200Hz. The data is subsequently packed into the standard starlink NDF format, which contain extensive header information for pipeline processing. In engineering mode, multiple MCE can also be run from the external sync box clock, however the control of the MCE is independent of the telescope system and data is saved in a simple text based data frame format with headers added by the control scripts. The clock rate of the MCE (from internal or external clock) is 50MHz. We currently operate with a multiplexing frame rate of 9.53 kHz. That is we multiplex 41 rows of SQ1s and stay on each row for 128 50MHz clock cycles. After a settling period, a calculated SQ1 feedback is applied based on the previous frame measurement and then after a further waiting period, the output sampled and co-added for a further number of clock cycles. The SQ1 bias for that row is then turned off and the next energised and so on. The MCE has various data modes. We most commonly return filtered SQ1 feedback. In filtered mode, the MCE applies a digital 4-pole Butterworth low-pass filter to the SQ1 feedback for output, matched to the 200Hz science mode data output frame rate. The MCE filter response can be seen in figure 13. Using the four scripts described above, the properties of individual TES detectors or the whole sub-array can be mapped out for different TES bias, heater and temperature setting. There are many interlinked properties that determine the successful operation of a TES array. However for the operation of SCUBA-2 on the telescope, the key performance figures are the detector NEP and NEFD (noise equivalent flux density). These properties have some distribution for the detectors in each focal plane. The modal value and distribution of the NEP and NEFD directly affect the mapping speed of the instrument. 3.2 Additional heat load from the science grade sub-arrays The 850µm science grade sub-array was cooled down first in the Cardiff system. The fridge base temperature was much higher than expected. We estimated that the sub-array was contributing an additional 36µW heat load on the mixing chamber, compared to previous prototype sub-arrays modules. Thermal modelling indicated that a thick layer of GE varnish used to insulate the bare niobium-titanium wires in the thermal breaks was the likely cause. The 450µm science grade had a thinner layer of GE varnish. With the 450µm sub-array in the system, the fridge base temperature was close to 50mk but still elevated. Following testing in Cardiff the 850µm CEM was changed to match the 450µm. Installed in SCUBA-2, the instrument base temperature is similarly elevated and the MUX temperatures of both science grade subarrays are 15mk warmer than the installed commissioning grade arrays. The higher MUX temperature has a significant affect on performance, particularly for the 850µm sub-array. 3.3 850µm Tc Flattening  Figure 6. The results of Tc flattening on the 850µm sub-array; (Left) no flattening, (right) with flattening. The SQ1 bias current is used to selectively apply heat to the detectors, to smooth out the Tc variation.  Proc. of SPIE Vol. 7741 774106-8 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  The science grade 850µm detectors have a large spread in measured Tc, perhaps exaggerated by the excess heat leak from 1K causing uneven heating of the MUX wafer. This is a significant factor for setting up this sub-array. To smooth out the effect of the variation in Tc, a novel technique of ‘Tc flattening’ has been developed. By applying a higher SQ1 bias on selected number of rows for a few clock cycles at the start of the 128 cycles each row is on, the SQ1 are used as heaters. Tc flattening is row based, so it cannot correct for a Tc variation across the columns. For the 850 µm we are able to recover 10 or so rows (although not every column in the row), which would not otherwise have been in the transition, with no apparent adverse effects. The results of Tc flattening are shown in Figure 6. 3.4 Science grade sub-array detector characteristics and performance We describe the measured characteristics of typical science grade 450 and 850µm TES detectors. Table 1. Measured characteristics for science grade arrays. 450µm [r16c3]  850µm [r10c20]  Normal State Resistance  59.7m Ω  28.2m Ω  Total Power (Heater + IV)  274pW  43.2pW  Transition Temperature  194.8mk  135mk  Thermal conductance G (= dP/dT)  6.0nW/K  1.8nW/K  Phonon limited NEP  1.12x10-16 W/√Hz  4.2x10-17 W/√Hz  450µm  850µm  Figure 7. Measured transition temperatures (left) and thermal conductance (right), for the 450µm and 850µm science grade subarrays.  Proc. of SPIE Vol. 7741 774106-9 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  Transition temperatures and thermal conductance are shown in Figure 7. From these measurements the intrinsic phonon noise limited NEP =  2  4 K BT G , can be estimated (assuming a 1Hz bandwidth). The phonon noise limited NEP for the  450µm detectors are shown in Figure 8, together with a comparison of the phonon noise limited NEP and NEP measured from the detector current noise in the dark for the 850µm array.  450µm 850µm  Figure 8. (Left) Phonon noise limited NEP for the 450um science grade detectors. (Right) Histogram showing the 850µm phonon noise limited NEP (red) and a typical measured (from the current noise) NEP (blue) at the start of Science Observing. The insert shows the 850µm measured NEP detector map.  The 450µm science grade sub-array achieves the goal to have the intrinsic detector NEP at least a factor of two less than the background from the sky and telescope: 992 detectors (78%) have measured NEP less than 1.2e-16 W/rtHz. For the 850µm science grade sub-array, 703 detectors (55%) meet the more stringent requirements. However in practice (see Figure 8) the variation of Tc (particularly for the 850), compromises in the SQUID setup and external sources of noise picked up by the TES influence the array performance tremendously.  Figure 9. Current-voltage (IV) characteristics for a typical 850µm detector, at two heater powers are plotted. (Right) IV power is plotted against applied detector bias for a range of heater settings.  IV characteristics for a typical 850µmTES at different heater powers are shown in figure 9. From the IV curves and by assuming that the shunt resistance is 5m Ω (the design value), the TES resistance and IV power in the superconducting  Proc. of SPIE Vol. 7741 774106-10 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  transition cann be calculated d. An (almost)) constant pow wer plateau in the t transition iss observed in Figure F 9. The effective e heater resistaance per detector and any varriation across thhe array can bee calculated froom IV and pow wer curves for different heater settinggs. The normal state s resistancee of the 850µm m detectors on the science grrade array are less l than 40% of the expecteed values (see Fig 10), compared to th he commissionning grade 8500µm the 450µm m science gradee array. The tottal power handdling and intrinsic NEP P of the science grade 850 arrray are both too specification. The impact of o the reduced normal n state reesistance is to narrow the width of th he transition inn terms of appllied detector bias for each deetector. This coompounds the problem of the variattion in transitiion temperaturres across thee array and makes m it harderr to setup the whole sub-arrray into transition witth a single app plied bias and heater h setting. The T reduced detector resistannces values maaybe also push some of the detectorss into instabilitty, depending upon the effeective time connstant of the TES T and the strength s of anyy electro thermal feedbback.  Increas sing Heater  Figure 10. Thee electrical resisttance of an 850µ µm detector is pllotted against appplied detector biias current, for a series of heater powers. The TES resisttance is constantt in the normal state. s (Right) A similar s plot is shhown the commisssioning grade 850µm 8 sub-arrayy.  The small signal theory fo or voltage biased TES predicct a simple rellation betweenn the bias voltaage and the dcc current responsivity in the transitio on: s = -1/V.  Figure 11. NEP (blue) and currrent responsivityy (red) plotted aggainst the applieed TES bias voltage, with the heater power consttant. The solid red line is i a fit to the currrent responsivityy, for bias voltagges < 2.3µV.  For a fixed heater h power we w measured thhe relation betw ween the TES bias voltage and a the currentt responsivity and also the bias and the electrical NEP. N Figure 10 illustrates thhese relations for f a typical deetector on the 850µm 8 commiissioning grade array. We observe similar s results with the sciennce grade arrayys. The linearr fit of the ressponsivity agaainst bias voltage over the bias voltag ge range < 2.3µ µV, agrees welll with theory.  Proc. of SPIE Vol. 7741 774106-11 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  The measured NEP (from the mean detector current noise between 2 to 10Hz) starts to increase (degrade) as the responsivity falls below 1x106 A/W. At lower bias voltages the detector became unstable and above 2.3µV the detector is becoming mostly normal (ignoring hysteresis). One explanation for the rise in NEP as the responsivity reduces is the increasing dominance of noise from the SQUID multiplexer, which is higher on the commissioning grade array during this measurement.  Figure 12. Current responsivity and NEP are plotted against heater power, for a fixed detector bias. Two detector biases are shown. The solid lines are fits to the data. (The measurements are for a typical detector on the 850µm science grade sub-array).  For a fixed detector bias and varying the heater, we found that the responsivity increases with reduced heater power. The NEP is broadly constant (once the responsivity is > 1x106A/W) to lower heater power, despite the increasing responsivity. For the science grade 850µm detector (see Figure 12), a lower detector bias resulted in a higher responsivity, but also a lower constant NEP. As the heater is reduced further the responsivity suddenly collapses as the TES goes superconducting. One aspect that is worth noting from this figure: We show that the responsivity varies with power loading on the TES. This will include optical power. Therefore observing bright calibrator sources for example, such as Mars, may introduce a significant nonlinear response; similarly if the varying sky background is not corrected sufficiently or often enough with the heater, the responsivity may change during an observation. 3.5 Setup for optical testing and on-sky calibration To optimise the arrays for on-sky calibration and science observing the TES bias and heater settings are chosen to maximise the number of detectors that meet or exceed the NEP requirements. However out of spec detectors also contribute to the final maps. Therefore a more sophisticated optimisation is to weight all the working (i.e. in transition) detectors based on their NEP and optimise the TES bias and heater to maximise the mapping speed of the instrument. Data from unstable or variable detectors are thrown out by the data reduction pipeline but it was found best to turn these detectors off in advance via the setup, as flux jumps and out of lock elements with rapidly ramping SQ1 feedback can cause steps and increased noise on adjacent detectors. The optimised array setups for observing typically return 800 – 900 working (as determined by a linear response to a ramped signal on the heater, used to calibrate or flatfield) detectors at each wavelength with a spread or distribution of NEP as shown Figure 9 for the 850µm sub-array. The planned upgrade to the CEM is expected to increase the number and quality of working TES, particularly on the 850µm array. The effect of temperature oscillations in the dilution fridge on the output of the detectors is also expected to diminish. The 1K box is cooled directly by the dilution fridge still. The fridge oscillations are largest in the still and therefore are adversely coupled to the detector arrays via the CEM.  Proc. of SPIE Vol. 7741 774106-12 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  The low freqquency oscillattions from the fridge as seen by the detectorrs, present a chhallenge to the data reductionn team in making flat maps. m In Figurre 13 we show w a typical timee stream and spectrum s for ann 850µm TES in the dark. The T large fridge oscillaation has a perriod of approxiimately 30 secconds. Also visible in the spectrum is a thermal pickup from f the pulse tube cooolers (1.5Hz plus p higher harm monics). The frequenccy spectrums are a generally fllat from about 1Hz until the roll r off due to the MCE filterr (70Hz). How wever the low frequenccy pickup (from m the fridge) and a 1/f from thhe MUX and detectors, d togetther with any discrete d features in the frequency spectrum constraain the observinng modes for the t instrument.  Figure 13. (Leeft) A typical tim me stream and spectrum from an 850µm detectorr in the dark, shoowing periodic frridge oscillations. (Right) With thhe detector in th he normal state and a the output saampled at a higheer frame rate (1 kHz), k to show thhe MCE filter characteristics.  Figure 14. Dettector current forr a single 850µm m TES, scanningg over Mars. (Rigght) The resultannt scan map of Mars M at 850µm, for f a 60 second observation. The ampliitudes of the signnal are in units of o power (pW).  A short sectiion of a quick scan observation of Mars, output o from a single s TES at 850µm, is shoown in Figure 14. This observation highlights h the effect e of the frridge oscillatioons on the sciennce data. The mars m signals are a the large doownward ‘spikes’ in thhe detector currrent. For much less bright sources s the friddge temperaturre pickup dwarrfs the signal from f the source. One of the challen nges for the data d reduction is to remove this backgrounnd while preserving flux froom faint how in referencce [2]; we are making m progresss in this respeect. extended souurces. As we sh  Proc. of SPIE Vol. 7741 774106-13 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  4. CONCLUSION We have described the characterisation of the first science grade arrays for SCUBA-2. The intrinsic properties of the TES arrays at both wavelengths achieve the design requirements and goals for the instrument, with much higher detector ‘pixel’ yields than we achieved on prototype or commissioning grade arrays. However the performance of the science grade arrays in the instrument has been hampered by a number of issues which are principally thermal in nature. We expect an improvement in performance with the new cold electronics modules that will be installed as part of the array upgrade this summer. The optical performance of the detector arrays and the results of on-sky calibration are discussed separately.  REFERENCES [1] Craig, S. et al., “SCUBA-2: engineering and commissioning challenges of the world largest sub-mm instrument at the JCMT”, Proc. SPIE 7741, (2010). [2] Holland, W. S. et al., “SCUBA-2: first results and on-sky performance”, Proc. SPIE 7741, (2010). [3] Dempsey, J. T. et al., “Extinction correction and on-sky calibration of SCUBA-2”, Proc. SPIE 7741, (2010). [4] Audley, M. D., Duncan, W. D., Holland, W. S. et al., “Fabrication of the SCUBA-2 detector arrays”, Nuclear Instruments and Methods in Physics Research A 520, 483–486 (2004). [5] deKorte, P.A.J., Beyer, J., Deiker, S., Hilton, G.C., Irwin, K.D., et al., “Time-division superconducting quantum interference device multiplexer for transition-edge sensors”, Rev. Sci. Instrum. 74, 3807-3815 (2003) [6] Irwin. K.D., “An application of electrothermal feedback for high resolution cryogenic particle detection", Appl. Phys. Lett., 66(15), 1998-2000 (1995). [7] Irwin K. D. and Hilton G. C., [Cryogenic Particle Detection], C. Enss (Ed.) Topics Appl. Phys. 99, Springer-Verlag Berlin Heidelberg, 63-149 (2005). [8] Woodcraft, A. L., Ade, P.A.R., Bintley, D. et al., “Electrical and optical measurements on the first SCUBA-2 prototype 1280 pixel submillimetre superconducting bolometer array”, Review of Scientific Instruments, 78, 024502 (2007). [9] Gao, X., Kelly, D. et al., “Automatic setup of SCUBA-2 detector arrays”, Proc. SPIE, Vol. 7020, 702025 (2008) [10] Battistelli, E. S., Amiri, M., Burger, B. et al., “Functional Description of Read-out Electronics for Time-Domain Multiplexed Bolometers for Millimeter and Sub-millimeter Astronomy”, Journal of Low Temperature Physics, 151, 908914 (2008).  Proc. of SPIE Vol. 7741 774106-14 Downloaded from SPIE Digital Library on 13 Sep 2011 to Terms of Use: http://spiedl.org/terms  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items